Flipkart's endeavour to stay ahead of the curve has taken it to several interesting places. And since the year 2015, it has also got the e-commerce giant to the doorstep of academia.
Why, you ask?
Well, it is through collaboration that paths to new solutions are discovered. Like in 2020, when they joined hands with IIT Patna to develop machine translation capabilities intended to translate even user-generated content, which is much trickier than you would imagine. Since its colloquial with sometimes a dash of regional words as well.
And like this, the Made in India portal has ensured that even the students and institutes who have shaken hands with them are gaining learning and exposure, making it a two-way street all through. Mayur Datar, Chief Data Scientist at Flipkart tells us, right from the start, about how their journey with the leading institutes of our country started and how it has helped them and the students discover new horizons. Excerpts from a conversation:
In which year did this initiative commence and why did you feel the need to foster industry-academia partnerships?
India is a heterogeneous market and to cater to the unique needs of such a diverse customer base, businesses need to be backed by cutting-edge technology. As a homegrown e-commerce marketplace, Flipkart understands the importance of industry-academia collaboration in helping drive cutting-edge research and capability development. Since 2015, Flipkart has partnered with leading academic institutions across the country to drive research and development in various fields.
We are committed to building a ‘Culture of Innovation’ both within and outside of the organisation and have taken giant strides in joining hands with some of the leading institutions in the country to fuel innovation at the grassroots level. Along with building solutions for e-commerce, the collaboration also aims to familiarise the students with real-world industry nuances as well as provide a platform for faculty members to work closely on research projects.
Do share details on Flipkart’s various academic partnerships and areas of collaboration such as AI, natural language processing, social commerce and so on.
Through our collaborations with leading academic institutions across the country, we offer research grants and market insights to develop hands-on prototypes as well as publish research papers jointly with academia. Our aim is to establish deeper academic collaborations which could help students and the institution to leverage our data and platform knowledge to work on India-specific e-commerce challenges, in addition to publishing research papers.
Some of the areas of collaboration include machine translation, social commerce, trust/safety, product search, user-generated content, recommendations and cataloguing — leveraging machine learning applications of natural language processing, computer vision and so on.
So far, Flipkart has partnered with 10 academic institutions including the Indian Institute of Science (IISC), IITs (Patna, Kharagpur, Bombay and Kanpur) and IIMs (Ahmedabad and Kolkata) to accelerate research and solve some of the most critical industry problems.
How will the students benefit through these partnerships directly?
A healthy collaboration between industry and academia is essential to catalyse innovation and growth in technology. While the industry offers key market insights, data and computes resources, academia focuses on building new knowledge through research and imparting education to students, the combination of which yields accelerated development of new breakthroughs.
Such collaborations bring real-world industry exposure to students, scholars and faculty member(s)of institutions and provide an opportunity to leverage our platform knowledge to work on India-specific e-commerce challenges. Some of the areas of collaboration include joint research activities, writing research papers, organising seminars and internship/mentorship opportunities.
Over the past years, nearly 60 students and professors have been involved with Flipkart under various academic partnerships and this has resulted in 33 research papers published in reputed conferences.
Highlight a few successful collaborations the company witnessed and the outcomes they have generated.
At Flipkart, our efforts have been focused on delivering an inclusive e-commerce experience for customers across the country. We think that the future is in offering customers a truly personalised shopping experience in the entire value chain, from search to after-sales support. Part of this is innovating in the voice and vernacular spaces, allowing online shoppers to interact with the platform in their own language.
Over the past three years, we have introduced 11 Indian language interfaces including Hindi, Bengali, Tamil, Gujarati, Odia, Malayalam and so on, by translating over 5.4 million words for each language.
In 2020, we partnered with IIT Patna to build machine translation capabilities for translating user-generated content. Building translation models for this content was particularly challenging as the content is dynamic in nature and needed to be translated while preserving the essence of the message.
For example, a product description by a seller may mostly have a predetermined set of keywords or a similar style format while a user review can include colloquial usages, mix of English and vernacular language and may even have incorrect usages. There were no predefined models that could have helped us translate this dynamic user-generated content. As we were working with Indic languages, there were no prior data models available for product reviews either that we could use for reference.
We worked with IIT Patna to build translation models from scratch by borrowing data sets from other non e-commerce domains such as news and legal to build prototype models while utilising Flipkart’s understanding of our customer base. These models were synthetic in nature and we worked on tailoring them for a product review by refining the data. We introduced variations into the data such as formal and informal styles, thereby, further refining it to identify variations in user-generated content.
While we started our collaboration by translating user reviews from English to Hindi and similar languages like Bengali, now we have expanded the translations to Dravidian languages such as Tamil.
How do academic partnerships play a pivotal role in building innovative solutions that lead to ecosystem development?
As India progresses on its digitalisation journey, e-commerce has been at the forefront of driving digital adoption through local innovation and technology development, making commerce seamless. As millions adopt e-commerce, there is a need to innovate at an ecosystem level to create an active flow of new technological developments to sustain this growth and provide a seamless experience for new and existing customers. Industry-academia collaborations can drive constructive research that translates into new solutions which can help achieve ecosystem benefits.
Moreover, industry-academia collaboration offers an opportunity for students to get involved in market-linked research and capability development to solve real-world problem statements. While the industry offers key market insights, academia focuses on building new knowledge through research and imparting education to students, the combination of which yields accelerated development of new breakthroughs
With industry-applied research and harnessing young talent, Flipkart’s academic collaborations help build critical solutions for e-commerce challenges. This joint endeavour is essential to develop the ecosystem and ultimately make e-commerce more seamless and inclusive for users across the country, especially for customers beyond the metros.
Is the company working on any more such partnerships in the near future?
Earlier this year, we partnered with IIT Delhi’s unit FITT (Foundation for Innovation and Technology Transfer). As part of the MoU, FITT will work in collaboration with the teams at Flipkart in the areas of social commerce, product recommendation to resellers and the consumer product exploration journey.
We will continue to collaborate with academia to decode e-commerce problem statements and work closely on research projects aimed at solving relevant technical challenges while imparting real-world industry exposure to students.
